Multi-Scale Modeling Mechanical

Research Network

Fluid Sub-section

Reaz Shafqat
L: 3 T:2
 What is Fluid Mechanics
• Study of fluid at rest or at motion.
Fluid Mechanics
Fluid Statics Fluid Dynamics
Fluid body at motion.
Fluid body at rest. • Pipe flow, open channel flow.
• Hydrostatic Pressure, Surface tension. • Flow across a cylinder, vehicle (air
• Body immersed in fluid. craft, motor vehicle, water vehicles).
• Motion of fluid body container. • Air flow (weather pattern).
• …………etc. • Ventilation, Heating-Cooling systems.
• What not……!
 Analysis
There are 2 (canonical/ generalized) types of description of a fluid.
Fluid Mechanics
Discrete approach Continuum approach

Lagrangian Picture Eulerian Picture

 Discrete approach:
Fluid is made up of discrete particles, molecules. Fluid is
viewed as from one reference point.

I pick this particle and I observe its movement. Governing Eqns.

• Lagrangian invariant.
So, for a given initial position, 𝓛 = σ 𝑲. 𝑬 − 𝑷. 𝑬 = 𝒄𝒐𝒏𝒔𝒕𝒂𝒏𝒕
Future positions, velocities, accelerations are • Euler-Lagrangian equation of motion.
functions of ‘time’ only. Conservation laws can 𝝏𝓛 𝒅 𝝏𝓛
• 𝝏𝒙 − 𝒅𝒕 𝝏𝒙ሶ = 𝟎
be directly applied.
• Least Action Principle.
So, here you observe your ensemble evolve • Boltzmann distribution.
over time, which is a quite daunting task.
 Continuum approach
Fluid is a single continuous body on a certain region of space (control volume) and velocity,
pressure and temperature (if you consider heat transfer or compressible flow) is distributed
as fields.

Space (x,y,z) and time (t) are mainly

independent, but related implicitly.
velocities, pressure, density and other variables
෡
𝑽(x,y,z,t) = u(x,y,z,t) 𝒊Ƹ + v(x,y,z,t) 𝒋Ƹ + w(x,y,z,t) 𝒌 are functions of ‘position and time’ only.
P = P(x,y,z,t) Conservation laws cannot be directly applied.
You need Reynolds transport theorem.
Discrete approach
In discrete approach,
𝒅 𝝏
𝒂= 𝑽 = 𝑽
𝒅𝒕 𝝏𝒕
The time derivative is a linear operator.

Continuum approach
In continuum approach,
𝒅 𝝏
𝒂 = 𝒅𝒕 𝑽(𝒙, 𝒚, 𝒛, 𝒕) = 𝝏𝒕 𝑽(𝒙, 𝒚, 𝒛, 𝒕) + 𝑽 . 𝜵 (𝑽(𝒙, 𝒚, 𝒛, 𝒕))
The time derivative is a non-linear operator.

Transient term Advection term

𝝏𝓛 𝒅 𝝏𝓛 Euler-lagrange Eqn. you can apply the above materialistic
− =𝟎 𝒅 𝝏
𝝏𝒙 𝒅𝒕 𝝏𝒙ሶ
derivative ( 𝒅𝒕 = 𝝏𝒕 + 𝑽 . 𝜵 )
which sets up the foundation of lagrangian mechanics is also
heavily used in lagranian version of fluid mechanics.

For, practical purposes, we’ll stick to Eulerian version of fluid mechanics very rarely
lagrange approach is preferred.
 Continuum approach
Velocity field distribution, is varying with position but it may or may not vary with time.

This car is moving at a steady speed.

𝒅 𝝏
𝒂 = 𝒅𝒕 𝑽(𝒙, 𝒚, 𝒛, 𝒕) = 𝑽(𝒙, 𝒚, 𝒛, 𝒕) + 𝑽 . 𝜵 (𝑽(𝒙, 𝒚, 𝒛, 𝒕))
𝝏𝒕

This term is non-zero. This term is zero. This term is non-zero.

𝝏
This phenomena is known as steady state (𝝏𝒕 anything = 0).
 Reynolds Transport Theorem
• Whenever you say conservation law you say a system where, something is conserved.
• But in Euler approach we focus on fixed region of space where the fluid abides in over
the fluid system.
Bsys= some property of the system
we are interested in (Mass,
Momentum, Energy…etc)
b= Bsys / msys

Now,

ෝ 𝒅𝑨 = 𝑽. 𝜵 𝒅𝑽 (Green Gauss function)

𝑽. 𝒏
B = ‫𝑽𝒅 𝒃𝝆 ׬‬
𝑫 𝝏
න 𝝆𝒃 𝒅𝑽 = න (𝝆𝒃) 𝒅𝑽 + න 𝑽. 𝜵(𝝆𝒃) 𝒅𝑽
𝑫𝒕 𝝏𝒕 Property Property
Property
𝑫 𝝏 change change at
change of
න 𝝆𝒃 𝒅𝑽 = න{ (𝝆𝒃) + 𝑽. 𝜵(𝝆𝒃)}𝒅𝑽 inside the inlet
𝑫𝒕 𝝏𝒕 a system
control outlet.
with time.
Materialistic Derivative volume.
 Reynolds Transport Theorem (Incompressible Flow)
For, B = Mass
(Mass is conserved)
𝜵. 𝑽 = 𝟎 ( Incompressible flow, where density is constant)

For B= momentum
= Mass. 𝑽𝒆𝒍𝒐𝒄𝒊𝒕𝒚

Normally, gx and gz
are 0 and gy = g =
9.81 m/s2 (unless,
you are studying
multi-physics (plasma
physics, bubble (For incompressible flow)
chamber, electrolysis
in battery) Newton’s second law
Body Force Force
(force is conserved)
force Due to Due to
Change viscosity
In
Pressure
 Governing parameters
Dimensional:
Time (t), Space (𝑋), Density (𝜌), Pressure (P), Velocity (𝑉), Viscosity (𝜇).

Governing Eqns.
 Governing parameters
Non-Dimensional:
Reynold’s Number (Re), Euler’s Number (Eu), Grashoff’s Number (Gr).

They tell about some characteristics of the flow.

For, pipe flow Re < 2300

Re > 10,000
Governing parameters
 Governing parameters
𝝏𝒖 𝝏𝒗 𝝏𝒘
𝜵. 𝑽 = 𝟎 𝑜𝑟 + + =𝟎
𝝏𝒙 𝝏𝒚 𝝏𝒛
x= L X, y= L Y, z= L Z
And, u = 𝑈𝑟𝑒𝑓 𝑈 , v = 𝑈𝑟𝑒𝑓 𝑉 , w = 𝑈𝑟𝑒𝑓 𝑊

𝝏𝑼 𝝏𝑽 𝝏𝑾
+ + =𝟎
𝝏𝑿 𝝏𝒀 𝝏𝒁

Non dimensionalized Governing Eqns.

Now, we have the eqns, with given initial conditions,

𝑽 = 𝑽𝒐 𝝐 ℝ𝟑 and 𝑷 = 𝑷𝟎 𝝐 ℝ at t = 0

Lets solve the eqns. For 𝑽 and P for t > 0

But

Wait….!
So, the equations does not always provide smooth and unique solutions. Solutions doesn’t
always exist. Soltions have weaker convergance.
 Navier
Stokes

Solutions doesnot exist

Navier
Stokes
Solving Eqns with same initial conditions and getting different results means solution is not
unique.
 Navier
Stokes
Results are not well behaved, will have finite time blow up (inifinite velocity) .

Navier
Stokes
Tiny change in inlet results in enormous change at the outlet .
 Reason behind Un-solvability

The reason is turbulence. In turbulence flow is much random and chaotic difficult to
predict for future time.

(The evil one)

If the evil one is negligible the eqns can easily be solved.
• For, slow moving fluid, the eqns are solvable.
• For, 2 D the solutions exists.
Velocity is random, so, people tend to use average version of the Navier Stokes.

𝑉 = 𝑉ത + 𝑉ሖ

Velocity Averaged Velocity

Velocity Deviation

Terrance Tao
Smooth and unique solutions does not exist for averaged Navier
Stokes.

ALAS!
As engineers what do we care?
Even though we don’t care but global existence and smoothness of Navier
Stokes (Or, maybe),

But our goal is to make devices using these eqns.

“But didn’t you say eqns are not solvable all the time”

There is always numerical.

Computational Fluid Dynamics
Study of Fluid Mechanics using Linear Algebra (functional Analysis), Numerical techniques
and other Algorithms targeting specific phenomena.

Algorithm Numerical Techniques

SIMPLE (steady state) Numerical Differentiation

PISO (transient state) Numerical Integration
K- 𝝎 𝒎𝒐𝒅𝒆𝒍 (laminar model) RK Method
K - 𝜺 𝒎𝒐𝒅𝒆𝒍 (turbulent model)

functional Analysis

Banach fixed Point Theorem

 Modelling

We want to anlyze this phenomena.

𝝏
This is a steady state phenomena. ( anything = 0)
𝝏𝒕

𝝏𝒖 𝝏𝒗 𝝏𝒘
+ + =𝟎 Continuity
𝝏𝒙 𝝏𝒚 𝝏𝒛

𝑽. 𝜵 𝑽 = − 𝜵 𝑷 + 𝝁 𝜵𝟐 𝑽 Momentum
 The flow is one directional.
v ≈ 0 and w ≈ 0
𝝏𝒖 Continuity
=𝟎
𝝏𝒙

𝝏𝒖 𝝏𝑷 𝒅𝟐 𝒖 𝒅𝟐 𝒖 𝒅𝟐 𝒖 Momentum
𝐮. 𝝏𝒙 = − 𝝏𝒙 + 𝝁( + 𝒅𝒚𝟐 + 𝒅𝒛𝟐 )
𝒅𝒙𝟐

𝜕𝑢 𝑑2 𝑢
= 0, =0
𝜕𝑥 𝑑𝑥 2

𝒅𝟐 𝒖 𝒅𝟐 𝒖
and both are non zeros (For 3-D).
𝒅𝒚𝟐 𝒅𝒛𝟐
𝒅𝟐 𝒖
= 𝟎 for (2-D).
𝒅𝒛𝟐

𝝏𝒖 Boundary conditions (2 required),

=𝟎
𝝏𝒙
𝝏𝑷 𝒅𝟐 𝒖 𝒅𝟐 𝒖
u = 𝒖𝒐 , 𝑎𝑡 𝑥 = 0(inlet).
0= − 𝝏𝒙 + 𝝁 (𝒅𝒚𝟐 + 𝒅𝒛𝟐 ) (Poission eqn.)
u= 0, at the wall (no slip condition).
 Modelling
For laminar flow the solutions exists as in form of poission eqn.
For turbulent flow the scenario is much more complicated.

For both simple and complicated senarios,

Ansys FLUENT is there to save the day.

 Assignment
outlet
Simulate this shape in 3D using Ansys Fluent. Y

X
Z

Keynotes:
• Use laminar model (not K-𝜔 model).
• Include gravity.